I am running an off-grid installation with Multiplus II 48/5000, MPPT chargers and AC out coupled PV, using PV Assistant.
The AC coupled PV inverter is a SMA Sunny Boy 3000-TL21. Its frequency response is a linear reduction in output between 51Hz and 52Hz. These are the parameters I gave when setting up the ESS Assistant. See second screenshot.
The first screenshot shows what happens on a typical day with low AC loads when the battery reaches the target charge voltage of 55.2V.
Starting at 9:59 battery charge current is gradually reduced by reducing power from the MPPT chargers. Battery voltage remains almost constant just a tad over 55.2V during this time, as it should.
Then once the MPPTs are at zero output, from about 10:01 to 10:03 charge current remains constant (sourced from the AC coupled PV). That leads to an overshooting in battery voltage all the way to 55.6V. During that time the Mutliplus slowly increases frequency from 50Hz to 51Hz.
The Multiplus knows that any frequency changes between 50 and 51Hz will have no effect. When setting up the PV assistant it was told: output reduction starts at 51Hz. This is as designed, no fault of the Sunny Boy.
Once the frequency gets over 51Hz the AC inverter output is reduced and the battery voltage slowly returns to normal.
If I was charging to 3.6V cell voltage (57.6V battery voltage, as often is done for LFP), this same overshooting would trip the BMS, causing a completely unnecessary power outage. It is only my conservative 3.45V target cell voltage that leaves enough buffer.
This is a major bug in the PV assistant. Once the MPPT chargers are effectively turned off, with still a surplus of charge current, the Multiplus frequency should instantly jump to 51Hz and then slowly rise from there. That would cut out the 3 minutes of constant current that cause the over-voltage.
There is absolutely no benefit in slowly moving from 50Hz to 51Hz.
Interesting enough: when the battery target-voltage changes to float (53.6V) at about 9:08, the frequency does jump to 52Hz a lot quicker, reducing output from the AC coupled inverter to zero in an instant. That actually is a nuisance, since there is no need to rush to get to float.
It is worth mentioning that the target voltage is set by the BMS via CAN. A couple of reasons to use close-coupled communication. First: seeing things like this, I actually do trust the BMS to do a better job in protecting the battery than the Victron ecosystem. But also: the BMS allows a much finer grained adjustment, i.e. just 18 minutes absorption, while Victron would force a 1 hour minimum on the battery.
Edit: why are the MPPT controllers tuned down to zero before the frequency response happens on the AC side? Would it not make sense to ramp both of them down at the same time, or even to start ramping down the AC coupled PV first, to reduce the load going through the Multiplus?
p.s.: ESS Assistant has the same bug, which has exactly the same effect when the grid is disconnected.